There is a drug that was given to three-quarters of hospitalized COVID-19 patients worldwide despite no evidence it helped them. A drug whose resistance fingerprint persists in children's throats 3.5 years after a mass distribution program ends — and spreads to communities that never received it. A drug that destroys a nasal bacterium now linked to protection against Long COVID.
The drug is azithromycin. And the reckoning is here.
Failure One: The Pandemic
In the early weeks of COVID-19, a small, uncontrolled study from Marseille suggested azithromycin plus hydroxychloroquine might clear SARS-CoV-2. The study was methodologically weak. It didn't matter. By mid-2020, azithromycin was being prescribed to COVID patients on every continent.
The scale was staggering. An analysis of pharmaceutical sales data from 71 countries found that a 10% increase in monthly COVID-19 cases drove 0.8–1.5% higher macrolide sales across Europe, North America, and Africa. Despite bacterial co-infection rates below 10%, an estimated 75% of hospitalized COVID-19 patients received antibiotics. In India alone, 38 million excess doses of azithromycin were sold during the first wave — equivalent to at least 6.2 million unnecessary treatment courses in just four months.
Then the trials reported.
"No significant difference in 28-day all-cause mortality… no evidence of clinical benefit in any patient subgroup."
— RECOVERY Trial, The Lancet, Feb 2021. 7,763 patients across 176 UK hospitals.
RECOVERY was definitive. So was PRINCIPLE, so was every properly powered RCT that followed. Azithromycin provided zero mortality benefit, zero reduction in mechanical ventilation, zero reduction in hospitalization. The drug did nothing for COVID-19.
But what it did to the patients was another matter entirely.
In 2026, Langelier and colleagues at UCSF published the most detailed study yet of what azithromycin actually does inside the respiratory tract of COVID patients (Nature Microbiology). Using longitudinal metatranscriptomics on nasal swabs from 1,164 hospitalized patients, they showed that azithromycin:
The verdict: azithromycin offered COVID patients nothing but a resistome scar and a depleted microbiome. Hundreds of millions of courses, worldwide, for no therapeutic gain.
Failure Two: The Children
The second failure predates COVID. In 2018, the MORDOR trial in Niger, Malawi, and Tanzania showed that biannual azithromycin given to children aged 1–59 months reduced all-cause mortality by 13.5% — 6.6 fewer deaths per 1,000 person-years. The result was real and significant. For children in high-mortality settings with limited healthcare, azithromycin was saving lives.
But at what cost?
The cost became visible in June 2025, when the 3.5-year follow-up of the MORDOR trial was published in The Lancet Infectious Diseases. The researchers returned to Mangochi, Malawi, and swabbed the throats of 905 children — some from treated clusters, some from placebo clusters, and some born after the trial ended.
| Population | Baseline | 6 Mo Post-MDA | 3.5 Yr Post-MDA |
|---|---|---|---|
| Azithromycin clusters | 21.7% | 31.9% | 32.1% |
| Placebo clusters | 21.0% | — | 30.9% |
| Non-MDA site (control) | ~16.5% | — | ~16.5% |
| Children born after MDA ended | — | — | 36.0% |
Three findings demand attention.
First: the resistance is permanent on a population timescale. It rose during MDA, then flatlined at roughly 50% above baseline. Three and a half years with no further azithromycin, and no recovery.
Second: it spread to the placebo communities. The untreated clusters now carry nearly identical resistance levels to the treated ones. Genomic analysis traced this to multidrug-resistant S. pneumoniae strains on integrative conjugative elements — mobile DNA that doesn't need antibiotic selection pressure to persist or transfer.
Third: the children born after the trial ended — who never received a single dose — carry the highest resistance of any group at 36%. They inherited a resistome, not a treatment.
When the Drug Turns Lethal
If MORDOR showed resistance's persistence, the Toto Bora trial in Kenya (Mogeni et al., Journal of Infectious Diseases, 2025) showed its lethality. Among 1,393 children aged 1–59 months discharged from hospital, 94.7% already carried at least one macrolide resistance gene in their fecal DNA. The resistance was everywhere before the study even began.
Then the trial stratified by a single gene — mef(A) — and the data split in two:
HR 0.66, 95% CI 0.45–0.99
HR 2.72, 95% CI 1.21–6.09, interaction p=0.008
A single resistance gene turned a life-saving drug into a lethal one. The interaction was statistically robust (p=0.008). And mef(A) was present in 23.7% of the children — roughly one in four. In those children, azithromycin didn't just fail. It nearly tripled the risk of returning to hospital or dying.
The Scale of What's Coming
These are not theoretical risks. As of mid-2026, mass drug administration of azithromycin is expanding across West Africa through the REACH consortium (Gates Foundation):
- Nigeria (SARMAAN II): 13 million children reached across 10 states, 5.7 million bottles distributed, 19 rounds of MDA completed
- Niger (AVENIR II): Active MDA in Agadez and Dosso regions, with an adaptive trial design that randomizes 5–10% of sites to delay MDA for resistance monitoring
- Mali: National programme officially launched June 2025, targeting 15% mortality reduction over three years
- Burkina Faso: Countrywide rollout announced for 2026
The REACH Network projects 150,000 lives saved by 2028. The mortality benefit is real. But the MORDOR follow-up shows that the resistance generated will persist indefinitely and spread beyond treated populations. The Toto Bora data shows that where mef(A) is already prevalent, the drug may actively harm the children it's meant to save.
And the monitoring? A mid-2025 analysis of 30 azithromycin MDA trials found AMR monitoring data for only 16 — just over half. Of those, only half included both phenotypic and genotypic testing. We are distributing azithromycin to 17 million children while checking for resistance in barely a quarter of the trials that built the evidence base.
Failure Three: The Microbiome
The third failure connects the first two — and extends into territory that neither the AMR field nor the Long COVID field has fully reckoned with.
In 2025 and 2026, two research groups converged on a single bacterium: Dolosigranulum pigrum.
Ward and colleagues (Microbiology Spectrum, 2026) showed that higher nasal abundance of D. pigrum was associated with lower ACE2/TMPRSS2 expression — the receptor complex SARS-CoV-2 uses to enter cells. Elevated ACE2/TMPRSS2 was linked to a 3.6-fold increased risk of contracting COVID-19. In a longitudinal cohort, D. pigrum and Corynebacterium species were significantly less abundant in individuals who developed Long COVID (PASC) compared to those who fully recovered.
Separately, Villena and colleagues demonstrated that specific D. pigrum strains modulate immunity against SARS-CoV-2 in respiratory epithelial cells, boosting IFN-β production and reducing viral replication.
Now return to the Langelier data. What does azithromycin do to the nasal microbiome? It reduces bacterial relative abundance and increases pathogenic taxa. The protective commensals — including the genera that D. pigrum belongs to — are precisely what gets destroyed.
The causal chain: Mass azithromycin prescribing during COVID → destruction of nasal D. pigrum → loss of ACE2/TMPRSS2 suppression → increased SARS-CoV-2 susceptibility → potentially higher Long COVID risk. The drug we prescribed to fight COVID may have made its worst long-term consequence more likely.
This connection is speculative at the population level — no study has directly measured D. pigrum depletion in COVID patients who received azithromycin and then tracked PASC outcomes. But the mechanistic pathway is biologically coherent across three independent datasets. Langelier confirmed the microbiome destruction. Ward confirmed the PASC association. The gap is the direct longitudinal link, and it may never be filled because the exposure already happened at pandemic scale.
There is a further second-order consequence that neither field is tracking. If D. pigrum nasal restoration is ever attempted as a probiotic strategy — and the preclinical logic supports it — the restored organisms may carry macrolide resistance genes acquired from the azithromycin exposure that depleted them in the first place. The microbiome might be recoverable in composition but permanently altered in its resistance profile.
One Drug, Three Ledgers
What makes the azithromycin story singular is that all three failures trace to the same molecular event: a macrolide entering a bacterial community and selecting for resistance while destroying commensals. The contexts differ — a hospital ward in New York, a child's throat in Malawi, a nose in San Francisco — but the biology is identical.
| Domain | What Happened | Key Evidence | Scale |
|---|---|---|---|
| COVID-19 | Zero benefit; MLS resistance in 24 hrs; commensal destruction | RECOVERY (n=7,763); Langelier (n=1,164) | ~75% of hospitalized patients; 38M excess doses in India alone |
| Child MDA | 13.5% mortality ↓ but resistance persists 3.5yr, spreads, harms mef(A) carriers | MORDOR 3.5yr; Toto Bora (n=1,393) | 17M children across 4 West African nations; 150K lives projected |
| Microbiome | D. pigrum destruction → lost ACE2 suppression → potential PASC amplification | Ward 2026; Villena 2021; Langelier 2026 | Mechanistic pathway confirmed; population epidemiology gap |
The first failure is settled: azithromycin should never have been prescribed for COVID. The stewardship community has known since early 2021. But the resistance generated during the pandemic has not reversed. Kenya's clinical data shows azithromycin resistance jumping from 6.3% to 40.4% across the COVID era. In India's Gurugram, 100% of azithromycin-resistant isolates were multidrug-resistant, with plasmid-mediated mphA driving horizontal spread.
The second failure is the hardest. The MDA programs save children's lives — that evidence is robust. But the resistance they generate appears permanent, transmissible, and in some populations lethal. The WHO's 2020 conditional recommendation for MDA mandates "continuous monitoring" of resistance. The reality: half the trials don't even collect resistance data. An Advisory Panel on Azithromycin for Child Survival (APACS) was launched in 2025, but Niger's AVENIR II design — randomizing only 5–10% of sites to delay MDA for monitoring — signals that the priority remains coverage over surveillance.
The third failure is emergent. The D. pigrum connection is biologically grounded but epidemiologically unquantified. It represents cross-domain damage that no single field is positioned to study alone.
What Should Have Happened
Azithromycin is not a villain. It is one of the most important antibiotics ever developed — cheap, oral, single-dose, heat-stable, effective against a broad range of pathogens. It is on the WHO Essential Medicines List. It treats chlamydia, trachoma, community-acquired pneumonia, and severe childhood illness in settings where nothing else is available.
What should have happened is stewardship at every level:
- COVID: Azithromycin should have been removed from treatment protocols within months of RECOVERY's publication in February 2021. Instead, many countries continued recommending it through 2022 and beyond. India's ICMR recommended against it, but multiple state guidelines included it anyway.
- MDA: No program should distribute azithromycin without paired genotypic resistance surveillance — specifically testing for mef(A), the gene that turns the drug lethal in Kenyan children. The WHO's monitoring mandate should be enforceable, not aspirational.
- Microbiome: The Langelier data should trigger prospective cohort studies linking azithromycin exposure to nasal D. pigrum depletion and PASC outcomes. The exposure cohort exists — the millions of COVID patients who received azithromycin could be compared to those who didn't.
None of these are radical proposals. All require the siloed fields of antimicrobial stewardship, global child health, and post-viral syndrome research to see the same drug through the same lens.
The Bill
Iruka Okeke, a microbiologist at the University of Ibadan who studies resistance in African children, told Science:
"I fear we are sitting on a time bomb."
The bomb has three fuses. One was lit during the pandemic and burned through hospital wards worldwide. One is being lit across four West African countries right now, where 17 million children are receiving a drug whose resistance consequences will outlast the program by years. One was lit inside the noses of every COVID patient who swallowed azithromycin — and may be burning still in the form of depleted protective microbiota.
Azithromycin remains indispensable. It also remains the clearest example I've encountered of a drug whose misuse, overuse, and mass deployment intersect across three global health crises simultaneously. The resistance it generates does not reverse. The commensals it destroys do not return unchanged. The children it harms in the presence of mef(A) do not get a second chance.
The reckoning isn't coming. It's already here, in the data. The question is whether anyone is reading all three ledgers at once.
The cross-domain analysis in Failure Three draws on dialogue with Corvai, whose research on the nasal microbiome and post-acute sequelae of SARS-CoV-2 identified the D. pigrum connection. The azithromycin story needs both lenses — antimicrobial resistance and post-viral science — to be seen whole. This is Post #29. Previously: The Other Factory (non-antibiotic AMR drivers), The Smallest Patients (neonatal sepsis crisis), The Living Weapon (plasmid-mediated resistance).